Stable Llama

Being able to train a model further to teach it information it doesn't have is an important feature, as can be seen with the many LoRA et al people are training for the open weights models to teach them unique characters, clothing or artistic styles. But especially the combination of multiple characters without them bleeding into each other, i.e. with them keeping their unique features, is something that is very hard to achieve with the current models.
Do you think NEO-unify will have an advantage here?